GPT-4o beaten!

Howdy, wizards.

🗳️ I’m testing a new format for this newsletter – hit reply with “better” or “worse” to vote.

Ps I just made my first public GPT – check it out in the newcomers list at the end of this newsletter.

‍

Dario’s Picks

The most important news stories in AI this week

1. Anthropic releases their best model to date, Claude 3.5 Sonnet. It surpasses GPT-4o on reasoning, code, math and knowledge. And it has an awesome new feature: Artifacts.‎

Why it matters  Claude just one-upped ChatGPT in performance, and people are finding productive use cases for Artifacts. Summary below.

‍

2. Illya Sutskever launches Safe Superintelligence. Ex-Chief Scientist at OpenAI is back with a new company focusing solely on, you guessed it, building safe superintelligence.

 Why it matters  Illya is one of the brightest minds alive according to many – and we’re lucky to have him working on perhaps one of the most difficult challenges of today.

‍

3. AI tools under fire for blatant plagiarism lately. Perplexity plagiarised Forbes earlier this month, and is apparently ignoring the Robots Exclusion Protocol (scraping content without permission). Luma is ripping of Pixar.

 Why it matters  AI plagiarism is a vicious cycle: Content creators lose revenues and won’t make quality content anymore, meaning AI companies lose original data to train on. A mutually beneficial business model is needed urgently.

‍

4. OpenAI acquires Rockset, and their team. Rockset is a real-time analytics database with world-class indexing and querying capabilities. They’ll integrate their tech to power retrieval infrastructure in OpenAI’s products.

 Why it matters  Better retrieval capabilities means better solutions for using our own data with AI. Will we soon get ChatGPT’s upgraded data analysis with access to our business databases in real-time? That’d be powerful.

‍

Tool picks

‍

Our previous top picks

‎ Text to speech  🥇Speechify

 Stems  🥇Fadr‎

 Academic research  🥇SciSpace

 Presentations  🥇Gamma

 Flashcards  🥇Gizmo

‍

See all our AI reviews →

‍

Behind the news

Anthropic releases their best model to date: Claude 3.5 Sonnet

‍

Anthropic just dropped Claude 3.5 Sonnet, the first version of their Claude 3.5 model series. It has multiple tricks up its sleeve:

• Surpasses GPT-4o on benchmarks for reasoning, code, math and knowledge.
• Improved vision capabilities, and is now significantly better at things like interpreting graphs and reading text in images.
• 2x speed boost versus Anthropic’s previous top-tier model, Claude 3 Opus.

Claude now also has an amazing new feature, Artifacts. It’s similar to Code Interpreter - it lets you view, edit and build things - but opens in a side panel within Claude.

The new model is available for free in Claude (web and iOS app), and for developers building with the model through their API, it’s just 20% of the price of Anthropic’s previous top-tier model.

‍

Why it matters  Anthropic keeps upping the game, seemingly with far less hype than what surrounds OpenAI and Google launches. I’m happy to see it, as they’re way more vocal about approaching AGI cautiously and investing in safety research. Who knows if they’re honest, but at least they keep the debate on security alive and kicking.

‍

GPTs

Newcomers on whatplugin.ai’s top 1,000 list of GPTs in the last week. How rankings work.

I just made my first public GPT, Add to Calendar. It’s a convenient way to create calendar events: describe an event (using text or voice) and you get a file that imports to your calendar when you open it. You can invite attendees, add a video conferencing link, etc. Tested it mostly on my Mac with iCal so far. Would love your feedback on it – just hit reply here and let me know how it’s working for you.

‍

‍

See all 1,000 GPTs →

‍

---

Affiliate disclosure: To cover the cost of my email software and the time I spend writing this newsletter, I sometimes link to products and other newsletters. Please assume these are affiliate links. If you choose to subscribe to a newsletter or buy a product through any of my links then THANK YOU – it will make it possible for me to continue to do this.